Internet Growth Analysis: Strategic & Product-Level Metrics, Marginal ROI, and Advertising Market Dynamics under Cost‑Reduction

DI (Data‑Information) is the essential bridge from BI (Business Intelligence) to AI, turning business data into knowledge graphs and enabling intelligent decision‑making.

The presentation is organized around four topics: (1) Internet growth analysis work, (2) Limitations of traditional ROI measurement, (3) Advantages of marginal ROI based on AB experiments, and (4) Advertising market competition under cost‑reduction.

1. Internet Growth Analysis

Strategic analysis (macro) covers corporate financial health, product health metrics, device LTV forecasts, and market size estimation to support top‑level decisions.

Product‑strategy analysis (micro) focuses on acquisition, retention, user‑experience optimization, and membership revenue strategies to guide product designers.

The core profit formula is Profit = (LTV - CAC) * Quantity , where LTV is lifetime value, CAC is customer acquisition cost, and Quantity is the number of users.

LTV estimation uses LTV(M) = Σ_{i=1}^{M} LT(i) * ARPU(i) , with LT(i) representing retention rate in month i and ARPU(i) the average revenue per user in month i.

Under cost‑reduction, the gross‑profit equations become:

Gross Profit = (LTV - CAC) * Quantity

Gross Profit = (ROI - 1) * CAC * Quantity

ROI = LTV / CAC

These highlight that ROI must exceed 1 for profitability.

2. Limitations of Traditional ROI

Traditional ROI (e.g., ROI = LTV / CAC for acquisition or ROI = GMV / TotalRecallCost for activation) cannot separate natural growth from campaign impact, leading to misleading conclusions, especially when seasonal effects cause ROI spikes.

Example: Experiment group 3 shows a high naive ROI (514%) but, after accounting for natural growth, its marginal ROI drops to 38%.

3. Marginal ROI Based on AB Experiments

Marginal ROI is calculated as:

Marginal ROI = (Avg(LTV_experiment) - Avg(LTV_control)) / (Avg(CAC_experiment) - Avg(CAC_control))

This removes the influence of the control group, revealing true incremental gains. In the case study, experiment group 2 achieves a marginal ROI of 145%, outperforming group 3's 38%.

Similarly, marginal CAC can be expressed as:

Marginal CAC = (TAC_experiment/UC_experiment - TAC_control/UC_control) / (NUC_experiment/UC_experiment - NUC_control/UC_control)

or, after scaling by the experiment pool size,

Marginal CAC = (TAC_experiment - TAC_control/UC_control * UC_experiment) / (NUC_experiment - NUC_control/UC_control * UC_experiment)

These formulas isolate the incremental acquisition cost of the experiment.

4. Advertising Market Competition under Cost‑Reduction

For advertisers, the shift from pure acquisition to profit‑preserving, fine‑grained growth makes marginal ROI a critical metric, and platforms need to provide AB‑testing capabilities to support it.

Platforms must also showcase natural growth contributions, protect effective traffic, and improve eCPM (revenue per thousand impressions) while fostering competitive bidding.

In summary, marginal ROI offers a more accurate assessment of advertising effectiveness in complex, cost‑constrained environments, benefiting both advertisers and platforms.